Browse / LLM wiki / LLM-Wiki-Setup
LLM-Wiki-Setup
alokm3hta-star
A living, grounded SAP knowledge base powered by a team of specialised AI agents running inside Claude Code.
Open alokm3hta-star/LLM-Wiki-Setup → Checked against the four rules.
Built for Claude Code
ships .claude CLAUDE.md and nothing for other agents.
In its own words
A living, grounded SAP knowledge base powered by a team of specialised AI agents running inside Claude Code. Feed it your documents; ask it anything; trust every answer because every answer is cited.
The Series
This project, and the thinking behind it, is documented in a five-part series:
- Part 1: I built an AI that refuses to guess
- Part 2: Seven AI agents and the harness that keeps them honest
- Part 3: Texting my SAP knowledge base on WhatsApp
- Part 4: I asked it for ABAP, and it refused to guess the BAPI
- Part 5: Did anyone ask Joule for on-premises? I built the add-on that does similar
Table of Contents
- What Is This?
- A Quick Word on Claude Code and AI Agents
- The Big Picture: How Everything Connects
- Meet the Agent Team
- The Full Pipeline: From PDF to Answered Query
- The Knowledge Clusters
- Sarah's Curation Workflow
- How Alex Manages Aaron and Adrian
- The Guardrail System
- The AI Harness
- The Scripts Engine
- Integrations
- Folder Structure
- Command Reference
- Getting Started
1. What Is This?
The LLM Wiki is a grounded knowledge base for SAP consulting. "Grounded" means it will only tell you things it has actually read; it never invents answers from the AI's general training data. Every response is cited to a specific page and section.
It works like a very well-organised team of consultants who have read all your documents, remember everything, never hallucinate, and can be questioned at any time.
Quoting 3% of the README. Read the rest on GitHub ↗
Read from the repository's git tree on 2026-08-17.